Figures for Paper
Anna-Leigh Brown
06/10/2020
iPSC splicing
Read in the splicing results returned by MAJIQ and make a volcano plot, only highlight genes of interest with a label.
junc_colors = c("#648FFF","#fe6101")
ipsc_splicing = fread("/Users/annaleigh/Documents/GitHub/unc13a_cryptic_splicing/data/ipsc_splicing_results.csv")
ipsc_splicing %>%
mutate(junction_name = glue::glue("{gene_name} - {paste_into_igv_junction}")) %>%
mutate(`Novel Junction` = de_novo_junctions == 0) %>%
mutate(log10_test_stat = -log10(1 - p_d_psi_0_10_per_lsv_junction)) %>%
mutate(log10_test_stat = ifelse(is.infinite(log10_test_stat), 16, log10_test_stat)) %>%
mutate(graph_alpha = ifelse(p_d_psi_0_10_per_lsv_junction > 0.9, 1, 0.2)) %>%
mutate(label_junction = case_when(gene_name %in% c("UNC13A",
"UNC13B","PFKP","SETD5",
"ATG4B","STMN2") &
p_d_psi_0_10_per_lsv_junction > 0.9 &
deltaPSI > 0 ~ junction_name,
T ~ "")) %>%
ggplot(aes(x = deltaPSI, y = log10_test_stat)) +
geom_point(aes(fill = `Novel Junction`,alpha = graph_alpha), pch = 21, size = 4) +
geom_text_repel(aes(label = label_junction), point.padding = 0.3,
min.segment.length = unit(0, 'lines')) +
geom_hline(yintercept = -log10(1 - .9)) +
guides(alpha = FALSE, size = FALSE) +
scale_fill_manual(values = junc_colors) +
theme(legend.position = 'top') +
ggpubr::theme_pubr() +
xlab("delta PSI") +
ylab("-Log 10 Test Statistic") +
theme(text = element_text(size = 18))
UNC13A Cryptic PSI across TDP-43 KD experiments
We have 2 PSI’s for the UNC13A cryptic, one which includes both the short and long form of the cryptic, and one which does not. While we’re not mentioning in figure 1, given that the longer form of the cryptic appears in control cerebellum, one could argue that that junction should not be included in the PSI calculation here for UNC13A cryptic per se. In practice, this makes very little difference in the conclusions made on the cell lines, so I’ve included both for completeness.
The ‘C/G’ tells which genotypes were supported by RNA-seq on rs12973192. The NB cell lines are het, as is the WTC11 cell line. SH-SY5Y cells are homozygote for the major allele. There was variability on the Klim hMN set on allelic expression.
rel_rna_cryptic_amount = fread("/Users/annaleigh/Documents/GitHub/unc13a_cryptic_splicing/data/kd_experiments_relative_rna_and_unc13a_cryptic_junction_counts.csv")
rel_rna_cryptic_amount[,cryptic_psi_full := ( UNC13A_3prime +
UNC13A_5prime + UNC13A_5prime_2 +
UNC13A_5prime_3) / (UNC13A_annotated + UNC13A_3prime +
UNC13A_5prime + UNC13A_5prime_2 +
UNC13A_5prime_3)]
rel_rna_cryptic_amount %>%
ggbarplot(,
x = "source",
add = c("mean_se","jitter"),
y = "cryptic_psi_full",
fill = 'condition',
color = 'condition',
position = position_dodge(0.8)) +
ggpubr::theme_pubr() +
scale_fill_manual(
values = c("#40B0A6","#E1BE6A")
) +
scale_color_manual(
values = c("#1C2617","#262114")
) +
scale_x_discrete(labels=c("shsy5y_normed" = "NB cells \n C/G",
"ipsc_normed_ward" = "iPSC Neurons \n C/G",
"shsy5y_normed_no_snp" = "SH-SY5Y \n C/C",
"klim_normed" = "Klim iPSC MN \n C/G + C/C",
"sh_dzap" = "Appocher \n NB cells \n C/G")) +
ylab("UNC13A Cryptic PSI") +
xlab("") +
guides(color = FALSE) 
rel_rna_cryptic_amount %>%
ggbarplot(,
x = "source",
add = c("mean_se","jitter"),
y = "cryptic_psi",
fill = 'condition',
color = 'condition',
position = position_dodge(0.8)) +
ggpubr::theme_pubr() +
scale_fill_manual(
values = c("#40B0A6","#E1BE6A")
) +
scale_color_manual(
values = c("#1C2617","#262114")
) +
scale_x_discrete(labels=c("shsy5y_normed" = "NB cells \n C/G",
"ipsc_normed_ward" = "iPSC Neurons \n C/G",
"shsy5y_normed_no_snp" = "SH-SY5Y \n C/C",
"klim_normed" = "Klim iPSC MN \n C/G + C/C",
"sh_dzap" = "Appocher \n NB cells \n C/G")) +
ylab("UNC13A Cryptic PSI") +
xlab("") +
guides(color = FALSE) 
Paient summary statistics
First let’s look at only including tissues with detectable stmn2 or UNC13A CE
clean_data_table = fread(file.path(here(),"data","nygc_junction_information.csv"))
clean_data_table = clean_data_table %>%
mutate(call = fct_relevel(call,
"C/C", "C/G", "G/G")) %>%
mutate(number_g_alleles = as.numeric(call) - 1) %>%
mutate(unc13a_cryptic_leaf_psi = ifelse(is.na(unc13a_cryptic_leaf_psi),0,unc13a_cryptic_leaf_psi)) %>%
mutate(stmn2_psi_groups = as.numeric(cut_interval(log10(stmn_2_cryptic_psi_leaf), n = 2))) %>%
mutate(stmn2_psi_groups = case_when(stmn2_psi_groups == 1 ~ " Low STMN2",
stmn2_psi_groups == 2 ~ "High STMN2",
TRUE ~ "No STMN2"))
print(glue::glue("Number of unique patients: {clean_data_table[,length(unique(participant_id))]}"))## Number of unique patients: 377print(glue::glue("Number of unique tissue samples: {clean_data_table[,length(unique(sample))]}"))## Number of unique tissue samples: 1349print("Patients Per Disease Category")## [1] "Patients Per Disease Category"clean_data_table[,length(unique(participant_id)),by = disease]## disease V1
## 1: ALS-FTD 23
## 2: ALS 193
## 3: Control 77
## 4: Other 11
## 5: FTD 61
## 6: ALS-AD 12print("Tissues Per Disease Category")## [1] "Tissues Per Disease Category"clean_data_table[,length(unique(sample)),by = disease]## disease V1
## 1: ALS-FTD 110
## 2: ALS 764
## 3: Control 199
## 4: Other 70
## 5: FTD 138
## 6: ALS-AD 68print("Number of patients per rs12973192 genotype")## [1] "Number of patients per rs12973192 genotype"clean_data_table[,length(unique(participant_id)),by = call]## call V1
## 1: C/C 166
## 2: G/G 58
## 3: C/G 153print("Number of tissues per disease")## [1] "Number of tissues per disease"clean_data_table[,.N,by = c("disease","tissue_clean")]## disease tissue_clean N
## 1: ALS-FTD Frontal_Cortex 22
## 2: ALS Frontal_Cortex 132
## 3: Control Frontal_Cortex 40
## 4: Other Frontal_Cortex 11
## 5: ALS-FTD Lumbar_Spinal_Cord 15
## 6: ALS Lumbar_Spinal_Cord 105
## 7: Control Lumbar_Spinal_Cord 33
## 8: Other Lumbar_Spinal_Cord 9
## 9: ALS-FTD Cervical_Spinal_Cord 14
## 10: ALS Cervical_Spinal_Cord 103
## 11: Control Cervical_Spinal_Cord 32
## 12: Other Cervical_Spinal_Cord 10
## 13: ALS-FTD Motor_Cortex 28
## 14: ALS Motor_Cortex 175
## 15: Control Motor_Cortex 23
## 16: Other Motor_Cortex 16
## 17: ALS-FTD Cerebellum 13
## 18: ALS Cerebellum 129
## 19: Control Cerebellum 28
## 20: Other Cerebellum 8
## 21: FTD Cerebellum 58
## 22: FTD Frontal_Cortex 45
## 23: ALS-AD Cerebellum 11
## 24: ALS-AD Motor_Cortex 13
## 25: ALS-AD Cervical_Spinal_Cord 10
## 26: ALS-AD Lumbar_Spinal_Cord 11
## 27: ALS-AD Frontal_Cortex 12
## 28: ALS-AD Occipital_Cortex 7
## 29: ALS-AD Thoracic_Spinal_Cord 4
## 30: ALS Occipital_Cortex 37
## 31: ALS Thoracic_Spinal_Cord 33
## 32: ALS-FTD Occipital_Cortex 6
## 33: Control Temporal_Cortex 23
## 34: ALS Temporal_Cortex 23
## 35: FTD Temporal_Cortex 35
## 36: Other Occipital_Cortex 7
## 37: Other Thoracic_Spinal_Cord 6
## 38: Control Thoracic_Spinal_Cord 5
## 39: Control Occipital_Cortex 5
## 40: ALS-FTD Thoracic_Spinal_Cord 5
## 41: ALS Hippocampus 27
## 42: ALS-FTD Hippocampus 7
## 43: Other Hippocampus 3
## 44: Control Hippocampus 10
## disease tissue_clean Nprint("Number of partcipants by mutation and disease")## [1] "Number of partcipants by mutation and disease"clean_data_table[,length(unique(participant_id)),by = c("mutations","disease")]## mutations disease V1
## 1: None ALS-FTD 13
## 2: None ALS 145
## 3: C9orf72 ALS-FTD 10
## 4: None Control 77
## 5: None Other 11
## 6: SOD1 ALS 8
## 7: OPTN ALS 4
## 8: C9orf72 ALS 32
## 9: MATR3 ALS 1
## 10: ANG ALS 1
## 11: C9orf72 FTD 12
## 12: None ALS-AD 11
## 13: None FTD 42
## 14: C9orf72 ALS-AD 1
## 15: TBK1 FTD 2
## 16: MAPT FTD 5
## 17: FUS ALS 2print(glue::glue("Number of patients per pathology:"))## Number of patients per pathology:clean_data_table[,length(unique(participant_id)),by = .(pathology)]## pathology V1
## 1: ALS-FTD 23
## 2: ALS 193
## 3: control 77
## 4: Other 11
## 5: 13
## 6: ALS-AD 12
## 7: FTD-TDP-A 24
## 8: FTD-TDP-B 3
## 9: FTD-TDP-C 9
## 10: FTD-TAU 7
## 11: FTD-FUS 5UNC13A cryptic is an event that is specific to TDP-43 proteinopathy
FTLD-non-TDP are those with TAU and FUS aggregates
Non-tdp ALS are those with SOD1 or FUS mutations. The n’s are quite low on this unfortunately, only 8 ALS with SOD1 and 2 with FUS mutations.
First we look at detection rate in tissues affected by TDP-43 proteinopathy, For FTLD this is frontal and temporal Cortices, and for ALS this is lumbar, cervical, and thoracic spinal cord samples as well as motor cortex. For controls we also take all 6 tissues, frontal,temporal,motor cortices and the lumbar, cervical, and thoracic spinal cords.
(As a side note, once we do this the number of ALS-non-TDP drops down to 6 (2 FUS) because the ALS sample tissues are not balanced and not every participant has samples in every tissue)
####Inclusion reads by if TDP-potential####
boxplot_table = clean_data_table %>%
mutate(disease_group2 = case_when(disease == "Control" ~ "Control",
pathology %in% c("FTD-TDP-A","FTD-TDP-B","FTD-TDP-C") ~ "FTLD-TDP",
pathology %in% c("FTD-TAU","FTD-FUS") ~ "FTLD \n non-TDP",
mutations %in% c("SOD1","FUS") ~ "ALS \n non-TDP",
T ~ "ALS-TDP")) %>%
mutate(across(UNC13A_3prime_leaf:UNC13A_annotated_leaf, ~ .x / library_size,.names = "{.col}_library_norm")) %>%
filter(!tissue_clean %in% c("Choroid","Liver")) %>%
dplyr::select(sample,participant_id,mutations,disease_group2,pathology,tissue_clean,contains("_library_norm")) %>%
melt() %>%
filter(grepl("_3prime|_5prime_",variable)) %>%
group_by(sample) %>%
mutate(inclusion_reads = sum(value)) %>%
ungroup() %>%
unique() %>%
mutate(junction_name = case_when(variable == "UNC13A_3prime_leaf_library_norm" ~ " Novel 3'",
variable == "UNC13A_5prime_1_leaf_library_norm" ~ " Short Novel 5'",
variable == "UNC13A_5prime_2_leaf_library_norm"~ "Long Novel 5'")) %>%
mutate(disease_tissue = case_when((grepl("FTLD",disease_group2) & grepl("Cortex",tissue_clean)) ~ T,
(grepl("ALS",disease_group2) & grepl("Cord|Motor",tissue_clean)) ~ T,
(grepl("Occipital",tissue_clean)) ~ F,
(grepl("Control",disease_group2) & grepl("Cord|Cortex",tissue_clean)) ~ T,
TRUE ~ F)) %>%
mutate(tissue_clean = gsub("_"," ",tissue_clean))Looking at disease tissue only, so just taking the cords in ALS and the frontal and temporal cortex of FTLD and then the cord and cortices in Controls.
boxplot_table %>%
filter(disease_tissue == T) %>%
mutate(detected = inclusion_reads > 0) %>%
dplyr::select(participant_id,disease_group2,detected) %>%
unique() %>%
group_by(disease_group2) %>%
mutate(n_sample = n_distinct(participant_id)) %>%
mutate(n_sample_detected = sum(detected)) %>%
dplyr::select(disease_group2,n_sample,n_sample_detected) %>%
unique() %>%
mutate(detection_rate = n_sample_detected / n_sample) %>%
mutate(disease_group2 = gsub("Control"," Control",disease_group2)) %>%
mutate(detection_name = glue::glue("{disease_group2} \n ( {n_sample} )")) %>%
ggplot() +
geom_col(aes(x = detection_name, y = detection_rate)) +
ggpubr::theme_pubr() +
scale_y_continuous(lim = c(0,1),labels = scales::percent) +
ylab("Percent of Patients \n UNC13A Cryptic Detected") +
theme(text = element_text(size = 18)) +
xlab("N individuals")
disease_comparisons = list( c("Control","ALS-TDP"),
c("Control","ALS \n non-TDP"),
c("Control","FTLD-TDP"),
c("Control","FTLD \n non-TDP" ))
table_to_test = boxplot_table %>%
filter(disease_tissue == T) %>%
group_by(disease_group2) %>%
mutate(n_sample = n_distinct(sample)) %>%
mutate(disease_group2 = gsub("Control"," Control",disease_group2)) %>%
mutate(detection_name = glue::glue("{disease_group2} \n ( {n_sample} )")) %>%
dplyr::select(detection_name,
participant_id,
inclusion_reads,
disease_group2) %>%
unique()
test_pair = pairwise.wilcox.test(table_to_test$inclusion_reads, table_to_test$detection_name,
p.adjust.method = "BH") %>% broom::tidy()
test_pair = test_pair %>%
mutate(p_value_draw = case_when(p.value < 0.0001~ "***",
p.value < 0.01 ~ "**",
p.value < 0.05 ~ "*",
TRUE ~ paste0("Adj. p-value \n",as.character(round(p.value,2))))) %>%
mutate(y.position = seq(0.25,by = 0.1,length.out = 9))
table_to_test %>%
ggplot(aes(x = detection_name, y = inclusion_reads * 10^6)) +
geom_boxplot() +
geom_jitter(height = 0) +
scale_y_log10() +
ggpubr::theme_pubr() +
ylab("UNC13A cryptic inclusion \n reads per million") +
theme(text = element_text(size = 18)) +
xlab("N samples") +
stat_pvalue_manual(test_pair %>% filter(p.value < 0.05),
label = "p_value_draw") +
stat_compare_means()
####Inclusion reads by if TDP-potential####
boxplot_table %>%
filter(disease_group2 %in% c("Control",
"FTLD \n non-TDP",
"FTLD-TDP")) %>%
filter(tissue_clean %in% c("Cerebellum","Frontal Cortex","Temporal Cortex")) %>%
dplyr::select(disease_group2,
tissue_clean,
junction_name,
value) %>%
unique() %>%
ggplot(aes(x = disease_group2,
y = value * 10^6,
fill = junction_name)) +
geom_boxplot(show.legend = F) +
geom_jitter(height = 0,show.legend = F) +
scale_y_log10() +
ggpubr::theme_pubr() +
facet_grid(vars(junction_name),vars(tissue_clean)) +
ylab("UNC13A cryptic \n reads per million") +
theme(text = element_text(size = 18)) +
xlab("") +
scale_fill_manual(values = colorblind_pal()(4)[2:4]) 
Splitting by the type of junction in ALS
####Inclusion reads by if TDP-potential####
boxplot_table %>%
filter(disease_group2 %in% c("Control",
"ALS \n non-TDP",
"ALS-TDP")) %>%
filter(tissue_clean %in% c("Cerebellum","Motor Cortex","Temporal Cortex") | grepl("Cord",tissue_clean)) %>%
dplyr::select(disease_group2,
tissue_clean,
junction_name,
value) %>%
unique() %>%
mutate(disease_group2 = fct_relevel(disease_group2,"Control","ALS \n non-TDP","ALS-TDP")) %>%
ggplot(aes(x = disease_group2,
y = value * 10^6,
fill = junction_name)) +
geom_boxplot(show.legend = F) +
geom_jitter(height = 0,show.legend = F) +
scale_y_log10() +
ggpubr::theme_pubr() +
facet_grid(vars(junction_name),vars(tissue_clean)) +
ylab("UNC13A cryptic \n reads per million") +
theme(text = element_text(size = 18)) +
xlab("") +
scale_fill_manual(values = colorblind_pal()(4)[2:4]) +
theme(axis.text.x = element_text(size = 14))
clean_data_table %>%
mutate(disease_group2 = case_when(disease == "Control" ~ "Control",
pathology %in% c("FTD-TDP-A","FTD-TDP-B","FTD-TDP-C") ~ "FTLD-TDP",
pathology %in% c("FTLD-TAU","FTLD-FUS") ~ "FTLD-non-TDP",
mutations %in% c("SOD1","FUS") ~ "ALS \n non-TDP",
T ~ "ALS-TDP")) %>%
filter(disease_group2 %in% c("ALS \n non-TDP","ALS-TDP")) %>%
mutate(across(UNC13A_3prime_leaf:UNC13A_annotated_leaf, ~ .x / library_size,.names = "{.col}_library_norm")) %>%
filter(!tissue_clean %in% c("Choroid","Liver")) %>%
dplyr::select(sample,disease_group2,tissue_clean,contains("_library_norm")) %>%
melt() %>%
filter(grepl("_3prime|_5prime_1",variable)) %>%
group_by(sample) %>%
mutate(inclusion_reads = sum(value)) %>%
ungroup() %>%
dplyr::select(-variable,-value) %>%
unique() %>%
ggplot(aes(x = disease_group2, y = inclusion_reads * 10^6)) +
geom_boxplot() +
geom_jitter(height = 0) +
facet_wrap(~tissue_clean,nrow = 2) +
scale_y_continuous(trans = 'log10') +
ggpubr::theme_pubr() +
ylab("UNC13A cryptic inclusion reads per million") 
Relationship between UNC13A and STMN2 cryptic PSI
We noted in the cell lines that there was a linear relationship between the amount of TARDBP KD efficiency and the amount of observed UNC13A cryptic. As a quantitative measure of TDP-43 proteinopathy does not exist for these patients, no ranked staining, we used the amount of STMN2 present in a tissue to rank the tissue by the amount of TDP-43 proteinopathy, as we’ve already shown you can use that as a measure. Here we show the relationship in samples with at least 10 STMN2 annotated junctions and tissues where the PSI of UNC13A cryptic was also detectable, as measured by STMN2 cryptic occurring in the tissue.
Scatter plot showing the relationship in non-logspace between STMN2 and UNC13A cryptic PSI
####scatter plot showing the correlation in non-log space in STMN2 and cryptic PSI####
clean_data_table %>%
filter(STMN2_annotated_leaf > 10) %>%
filter(stmn_2_cryptic_psi_leaf > 0 ) %>%
filter(unc13a_cryptic_leaf_psi > 0 ) %>%
group_by(call) %>%
ggpubr::ggscatter(.,
x = "stmn_2_cryptic_psi_leaf",
y = "unc13a_cryptic_leaf_psi",
add = "reg.line",
cor.coef = T) +
ylab("UNC13A Cryptic PSI ") +
xlab("STMN2 Cryptic PSI ") ## `geom_smooth()` using formula 'y ~ x'
####scatter plot showing the correlation in non-log space in STMN2 and cryptic PSI####
clean_data_table %>%
filter(STMN2_annotated_leaf > 10) %>%
filter(stmn_2_cryptic_psi_leaf > 0 ) %>%
filter(unc13a_cryptic_leaf_psi > 0 ) %>%
mutate(call = fct_relevel(call,"C/C", "C/G", "G/G")) %>%
group_by(call) %>%
ggpubr::ggscatter(.,
x = "stmn_2_cryptic_psi_leaf",
y = "unc13a_cryptic_leaf_psi",
add = "reg.line",
color = "call") +
ylab("UNC13A Cryptic PSI ") +
xlab("STMN2 Cryptic PSI ") +
stat_cor(aes(color = call),show.legend = FALSE) +
scale_color_manual(values = c("#88CCEE","#44AA99","#105e2a"))## `geom_smooth()` using formula 'y ~ x'
You can see a clear relationship, which appears to be modulated by the genotype of rs12973192. This lead us to see if genotype was actually predictive of the amount of UNC13A cryptic psi. Because PSI’s tend to be log linear, we chose to model UNC13A in the log10 + 1 (log10(1) is zero) and we selected for tissues which show TDP-43 pathology, as we assume this is an event which only begins to show clear after TDP-43 pathology arises.
Scatter plot showing the relationship in log + 1 between STMN2 and UNC13A cryptic PSI
clean_data_table %>%
filter(disease_tissue == T) %>%
filter(disease_group2 %in% c("FTLD-TDP","ALS-TDP")) %>%
mutate(call = fct_relevel(call,
"C/C", "C/G", "G/G")) %>%
filter(STMN2_annotated_leaf > 10) %>%
filter(stmn_2_cryptic_psi_leaf > 0 ) %>%
filter(unc13a_cryptic_leaf_psi > 0 ) %>%
mutate(log10_stmn_2_cryptic_psi_leaf_plusone = log10(stmn_2_cryptic_psi_leaf + 1)) %>%
mutate(log10_unc13a_cryptic_leaf_psi_plusone = log10(unc13a_cryptic_leaf_psi + 1)) %>%
group_by(call) %>%
ggpubr::ggscatter(.,
x = "log10_stmn_2_cryptic_psi_leaf_plusone",
y = "log10_unc13a_cryptic_leaf_psi_plusone",
color = 'call',fill = 'call',
add = "reg.line") +
ylab("Log10 UNC13A Cryptic PSI + 1") +
xlab("Log10 STMN2 Cryptic PSI + 1 ") +
scale_color_manual(values = c("#88CCEE","#44AA99","#105e2a")) +
scale_fill_manual(values = c("#88CCEE","#44AA99","#105e2a"))## `geom_smooth()` using formula 'y ~ x'
We can perform this analysis separated by tissue, but low numbers make the results more complicated to interpret. Additionally, the tissue only matters in so much as we know that certain tissues in certain diseases are affected by TDP-43 pathology, whereas others aren’t, tissue is not the most interesting or important variable, TDP-43 pathology is the question.
####scatter plot showing the correlation in non-log space in STMN2 and cryptic PSI####
clean_data_table %>%
filter(tissue_clean%in% c("Cervical_Spinal_Cord",
"Motor_Cortex",
"Frontal_Cortex","Temporal_Cortex")) %>%
mutate(call = fct_relevel(call,
"C/C", "C/G", "G/G")) %>%
filter(STMN2_annotated_leaf > 10) %>%
filter(cryptic_psi > 0 ) %>%
filter(stmn_2_cryptic_psi > 0 ) %>%
group_by(call) %>%
ggpubr::ggscatter(.,
x = "stmn_2_cryptic_psi",
y = "unc13a_cryptic_leaf_psi",
color = 'call',
add = "reg.line") +
ylab("UNC13A Cryptic PSI ") +
xlab("STMN2 Cryptic PSI ") +
scale_color_manual(values = c("#88CCEE","#44AA99","#105e2a")) +
facet_wrap(~tissue_clean)## `geom_smooth()` using formula 'y ~ x'
Modeling UNC13A cryptic using glms and mixed models
This brought us to the idea that we could even potentially model UNC13A cryptic pathology using genotype and the amount of TDP-43 pathology. We’ve coded the number of G alleles here as a number and are looking into tissue with TDP-43 pathology(measure with STMN2 cryptic PSI).
new_regress = clean_data_table %>%
mutate(call = fct_relevel(call,
"C/C", "C/G", "G/G")) %>%
filter(STMN2_annotated_leaf > 10) %>%
filter(stmn_2_cryptic_psi_leaf > 0 ) %>%
mutate(log10_stmn_2_cryptic_psi_leaf = log10(stmn_2_cryptic_psi_leaf)) %>%
mutate(log10_unc13a_cryptic_leaf_psi_plusone = log10(unc13a_cryptic_leaf_psi + 1)) %>%
mutate(log10_stmn_2_cryptic_psi_leaf_plusone = log10(stmn_2_cryptic_psi_leaf + 1))
another_model2 = lm(log10(unc13a_cryptic_leaf_psi + 1) ~
log10(stmn_2_cryptic_psi_leaf + 1) * number_g_alleles + pathology + tissue_clean,
new_regress)
the_pretty_plot = pretty_effect_plot(another_model2,p_value_cutoff = 1,"UNC13A Cryptic PSI") + ggpubr::theme_pubr()
print(the_pretty_plot)
We find that at a cutoff of 0.01 for significant effects, only the interaction between the number of G alleles and the amount of STMN2 pathology is predictive of UNC13A cryptic inclusion. the absolute amount of STMN2 pathology does not vary between genotypes.
genotype_comparisons = list(c("C/C", "C/G"), c("C/C", "G/G"), c("C/G", "G/G"))
ggboxplot(new_regress, x = "call", y = "log10_stmn_2_cryptic_psi_leaf",
color = "call", palette = "jco")+
stat_compare_means() +
stat_compare_means(comparisons = genotype_comparisons)
ggboxplot(new_regress, x = "call", y = "stmn_2_cryptic_psi_leaf",
color = "call", palette = "jco")+
stat_compare_means() +
stat_compare_means(comparisons = genotype_comparisons)
new_regress %>%
ggplot() +
stat_ecdf(aes(color = call, x = log10_stmn_2_cryptic_psi_leaf))
We see that detection rate by genotype varies.
overall_fisher = clean_data_table %>%
filter(disease_tissue == T) %>%
filter(disease_group2 %in% c("FTLD-TDP","ALS-TDP")) %>%
mutate(unc13a_detected = unc13a_cryptic_leaf_psi > 0) %>%
dplyr::select(participant_id,unc13a_detected,call) %>%
unique() %>%
group_by(call) %>%
mutate(n_sample = n_distinct(participant_id)) %>%
mutate(n_sample_detected = sum(unc13a_detected)) %>%
dplyr::select(call,n_sample,n_sample_detected) %>%
unique() %>%
mutate(n_non_detected = n_sample - n_sample_detected) %>%
dplyr::select(-n_sample)
over_p = overall_fisher %>%
column_to_rownames('call') %>%
fisher.test() %>%
broom::tidy() %>%
.$p.value
overall_fisher %>%
mutate(n_sample = n_non_detected + n_sample_detected) %>%
mutate(detection_rate = n_sample_detected / n_sample) %>%
mutate(detection_name = glue::glue("{call} \n ( {n_sample} )")) %>%
ggplot(aes(x = detection_name, y = detection_rate, fill = detection_name)) +
geom_col(show.legend = F) +
ggpubr::theme_pubr() +
scale_y_continuous(labels = scales::percent) +
ylab("% of TDP-43 Proteionopathy Patients \n UNC13A Cryptic Detected") +
theme(text = element_text(size = 18)) +
xlab("N individuals") +
scale_fill_manual(values = c("#88CCEE","#44AA99","#1B7739")) 
Although this difference is not significant, with the Fisher’s exact giving a p-value of 0.28.
Only looking at disease relevant tissue and in patients
detection_table = clean_data_table %>%
filter(disease_tissue == T) %>%
filter(disease_group2 %in% c("FTLD-TDP","ALS-TDP")) %>%
mutate(call = fct_relevel(call,
"C/C", "C/G", "G/G")) %>%
mutate(stmn2_psi_groups = as.numeric(cut_interval(log10(stmn_2_cryptic_psi), n = 2))) %>%
mutate(stmn2_psi_groups = ifelse(is.na(stmn2_psi_groups)," No STMN2",stmn2_psi_groups)) %>%
mutate(stmn2_psi_groups = case_when(stmn2_psi_groups == 1 ~ " Low STMN2",
stmn2_psi_groups == 2 ~ "High STMN2",
TRUE ~ stmn2_psi_groups)) %>%
mutate(unc13a_detected = unc13a_cryptic_leaf_psi > 0) %>%
group_by(call,unc13a_detected,stmn2_psi_groups) %>%
add_count(name = "genotype_detected") %>%
dplyr::select(call,unc13a_detected,genotype_detected,stmn2_psi_groups) %>%
unique() %>%
pivot_wider(names_from = "unc13a_detected",
values_from = "genotype_detected") %>%
dplyr::rename(unc_not_detected = `FALSE`, unc_cryptic_detected = `TRUE`) %>%
mutate(total_tissue = (unc_not_detected + unc_cryptic_detected)) %>%
mutate(detection_rate = unc_cryptic_detected / total_tissue) %>%
ungroup() %>%
mutate(stmn2_psi_groups = fct_relevel(stmn2_psi_groups, "No STMN2", after = 0)) %>%
mutate(detection_name = glue::glue("{call} \n ( {total_tissue} )"))
detection_table %>%
ggplot(aes(x = call, y = detection_rate,fill = call)) +
geom_col(show.legend = F,position = 'dodge2') +
facet_wrap(~stmn2_psi_groups) +
scale_fill_manual(values = c("#88CCEE","#44AA99","#1B7739")) +
scale_y_continuous(lim = c(0,0.5),labels = scales::percent) +
ylab("Percent of Samples \n UNC13A Cryptic Detected") +
theme(text = element_text(size = 18)) +
ggpubr::theme_pubr() +
geom_text(aes(label = total_tissue,y = 0),vjust = 1,size = 6) +
theme(text = element_text(size = 18)) +
ggpubr::theme_pubr() +
xlab("N Tissues") 
UNC13A cryptic PSI
clean_data_table %>%
filter(disease_tissue == T) %>%
filter(disease_group2 %in% c("FTLD-TDP","ALS-TDP")) %>%
mutate(call = fct_relevel(call,
"C/C", "C/G", "G/G")) %>%
filter(stmn2_annotated > 10) %>%
filter(cryptic_psi > 0) %>%
ggbarplot( x = "call",
y = "cryptic_psi",
color = 'call',
position = position_dodge(0.8),
add = c("mean_se","jitter")) +
ylab("UNC13A Cryptic PSI") +
scale_color_manual(values = c("#88CCEE","#44AA99","#105e2a")) +
theme(text = element_text(size = 20)) +
stat_compare_means(aes(group = call), label = "p.format") +
stat_compare_means(aes(group = call),comparisons = genotype_comparisons,label = "p.signif")
clean_data_table %>%
filter(disease_tissue == T) %>%
filter(disease_group2 %in% c("FTLD-TDP","ALS-TDP")) %>%
mutate(call = fct_relevel(call,
"C/C", "C/G", "G/G")) %>%
mutate(stmn2_psi_groups = as.numeric(cut_interval(log10(stmn_2_cryptic_psi), n = 2))) %>%
mutate(stmn2_psi_groups = ifelse(is.na(stmn2_psi_groups)," No STMN2",stmn2_psi_groups)) %>%
mutate(stmn2_psi_groups = case_when(stmn2_psi_groups == 1 ~ " Low STMN2",
stmn2_psi_groups == 2 ~ "High STMN2",
TRUE ~ stmn2_psi_groups)) %>%
filter(cryptic_psi > 0) %>%
ggbarplot( x = "call",
y = "cryptic_psi",
color = 'call',
facet.by = 'stmn2_psi_groups',
position = position_dodge(0.8),
add = c("mean_se","jitter")) +
ylab("UNC13A Cryptic PSI") +
scale_color_manual(values = c("#88CCEE","#44AA99","#105e2a")) +
theme(text = element_text(size = 20)) +
stat_compare_means(aes(group = call), label = "p.format") +
stat_compare_means(aes(group = call),comparisons = genotype_comparisons,label = "p.signif") +
xlab("")